Learning from histopathology images
Computational Pathology comprises automated cell and cell nucleus detection, segmentation and staining estimation. Modern random forest models are capable to learn from expert annotated images the differences between various kinds of cancer related outcomes.
We collaborate with outstanding institutions in several projects to learn how pathologists understand the images and which apects are important for reliable, repoducible and more objective diagnostics. Read more...
Collaboration: MSKCC, New York, USZ, Switzerland
How to learn from petabytes of data?
Deep learning experiences a tremendous rise in attention in the machine learning community. The increasing amount of available data allows for the learning of highly accurate patterns with various kinds of neural networks. We explore the potential of deep learning algorithms on histopathologic images for cancer diagnosis and grading, where conventional machine learning methods are usually limited. Read more...
Find Distinctive Patterns in High Dimensional Expression Profiles
Medical statistics and machine learning techniques are incorporated to find distinctive expression profiles in high dimensional genetic or proteomic medical data. Found biomarker candidates are then subject for validation on large patient cohorts. Read more...
Collaboration: MSKCC, USZ
Computational Radiology and Crohn's Disease
Can we detect and grade CD in MRI automatically?
Crohn's disease (CD) is a chronic inflammatory bowel disease which affects one of more parts of the terminal ileum and/or colon. Longterm monitoring of the disease is necessary for appropriate treatment and and also research. Automatic detection and scoring of CD with magnetic resonance imaging would substantially facilate and improve personalized and objective screening. Read more...